Changeset 8742 for branches/HeuristicLab.TimeSeries/HeuristicLab.Algorithms.DataAnalysis/3.4/SupportVectorMachine/SupportVectorRegression.cs

Timestamp:

10/05/12 11:58:17 (12 years ago)

Author:

mkommend

Message:

#1081: Merged trunk changes and fixed compilation errors due to the merge.

File:

• branches/HeuristicLab.TimeSeries/HeuristicLab.Algorithms.DataAnalysis/3.4/SupportVectorMachine/SupportVectorRegression.cs (modified) (9 diffs)

branches/HeuristicLab.TimeSeries/HeuristicLab.Algorithms.DataAnalysis/3.4/SupportVectorMachine/SupportVectorRegression.cs

@@ -30 +30 @@
 using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
 using HeuristicLab.Problems.DataAnalysis;
+using LibSVM;
 
 namespace HeuristicLab.Algorithms.DataAnalysis {
@@ -45 +46 @@
     private const string GammaParameterName = "Gamma";
     private const string EpsilonParameterName = "Epsilon";
+    private const string DegreeParameterName = "Degree";
 
     #region parameter properties
@@ -64 +66 @@
     public IValueParameter<DoubleValue> EpsilonParameter {
       get { return (IValueParameter<DoubleValue>)Parameters[EpsilonParameterName]; }
+    }
+    public IValueParameter<IntValue> DegreeParameter {
+      get { return (IValueParameter<IntValue>)Parameters[DegreeParameterName]; }
     }
     #endregion
@@ -86 +91 @@
     public DoubleValue Epsilon {
       get { return EpsilonParameter.Value; }
+    }
+    public IntValue Degree {
+      get { return DegreeParameter.Value; }
     }
     #endregion
@@ -111 +119 @@
       Parameters.Add(new ValueParameter<DoubleValue>(GammaParameterName, "The value of the gamma parameter in the kernel function.", new DoubleValue(1.0)));
       Parameters.Add(new ValueParameter<DoubleValue>(EpsilonParameterName, "The value of the epsilon parameter for epsilon-SVR.", new DoubleValue(0.1)));
+      Parameters.Add(new ValueParameter<IntValue>(DegreeParameterName, "The degree parameter for the polynomial kernel function.", new IntValue(3)));
     }
     [StorableHook(HookType.AfterDeserialization)]
-    private void AfterDeserialization() { }
+    private void AfterDeserialization() {
+      #region backwards compatibility (change with 3.4)
+      if (!Parameters.ContainsKey(DegreeParameterName))
+        Parameters.Add(new ValueParameter<IntValue>(DegreeParameterName, "The degree parameter for the polynomial kernel function.", new IntValue(3)));
+      #endregion
+    }
 
     public override IDeepCloneable Clone(Cloner cloner) {
@@ -126 +140 @@
       int nSv;
       var solution = CreateSupportVectorRegressionSolution(problemData, selectedInputVariables, SvmType.Value,
-        KernelType.Value, Cost.Value, Nu.Value, Gamma.Value, Epsilon.Value,
+        KernelType.Value, Cost.Value, Nu.Value, Gamma.Value, Epsilon.Value, Degree.Value,
         out trainR2, out testR2, out nSv);
 
@@ -136 +150 @@
 
     public static SupportVectorRegressionSolution CreateSupportVectorRegressionSolution(IRegressionProblemData problemData, IEnumerable<string> allowedInputVariables,
-      string svmType, string kernelType, double cost, double nu, double gamma, double epsilon,
+      string svmType, string kernelType, double cost, double nu, double gamma, double epsilon, int degree,
       out double trainingR2, out double testR2, out int nSv) {
       Dataset dataset = problemData.Dataset;
@@ -143 +157 @@
 
       //extract SVM parameters from scope and set them
-      SVM.Parameter parameter = new SVM.Parameter();
-      parameter.SvmType = (SVM.SvmType)Enum.Parse(typeof(SVM.SvmType), svmType, true);
-      parameter.KernelType = (SVM.KernelType)Enum.Parse(typeof(SVM.KernelType), kernelType, true);
+      svm_parameter parameter = new svm_parameter();
+      parameter.svm_type = GetSvmType(svmType);
+      parameter.kernel_type = GetKernelType(kernelType);
       parameter.C = cost;
-      parameter.Nu = nu;
-      parameter.Gamma = gamma;
-      parameter.P = epsilon;
-      parameter.CacheSize = 500;
-      parameter.Probability = false;
-
-
-      SVM.Problem problem = SupportVectorMachineUtil.CreateSvmProblem(dataset, targetVariable, allowedInputVariables, rows);
-      SVM.RangeTransform rangeTransform = SVM.RangeTransform.Compute(problem);
-      SVM.Problem scaledProblem = SVM.Scaling.Scale(rangeTransform, problem);
-      var svmModel = SVM.Training.Train(scaledProblem, parameter);
-      nSv = svmModel.SupportVectorCount;
+      parameter.nu = nu;
+      parameter.gamma = gamma;
+      parameter.p = epsilon;
+      parameter.cache_size = 500;
+      parameter.probability = 0;
+      parameter.eps = 0.001;
+      parameter.degree = degree;
+      parameter.shrinking = 1;
+      parameter.coef0 = 0;
+
+
+
+      svm_problem problem = SupportVectorMachineUtil.CreateSvmProblem(dataset, targetVariable, allowedInputVariables, rows);
+      RangeTransform rangeTransform = RangeTransform.Compute(problem);
+      svm_problem scaledProblem = rangeTransform.Scale(problem);
+      var svmModel = svm.svm_train(scaledProblem, parameter);
+      nSv = svmModel.SV.Length;
       var model = new SupportVectorMachineModel(svmModel, rangeTransform, targetVariable, allowedInputVariables);
       var solution = new SupportVectorRegressionSolution(model, (IRegressionProblemData)problemData.Clone());
@@ -165 +184 @@
       return solution;
     }
+
+    private static int GetSvmType(string svmType) {
+      if (svmType == "NU_SVR") return svm_parameter.NU_SVR;
+      if (svmType == "EPSILON_SVR") return svm_parameter.EPSILON_SVR;
+      throw new ArgumentException("Unknown SVM type");
+    }
+
+    private static int GetKernelType(string kernelType) {
+      if (kernelType == "LINEAR") return svm_parameter.LINEAR;
+      if (kernelType == "POLY") return svm_parameter.POLY;
+      if (kernelType == "SIGMOID") return svm_parameter.SIGMOID;
+      if (kernelType == "RBF") return svm_parameter.RBF;
+      throw new ArgumentException("Unknown kernel type");
+    }
     #endregion
   }